Artificial Intelligence1 min read

Build an On-Call Team of AI Agents on DigitalOcean MicroVMs

By · Published by Everything Blog

In short

DigitalOcean's incident response architecture divides responsibilities among six specialized agents covering commander, investigator, remediator, and verifier roles. The system executes diagnostics across commits, metrics, and snapshots in parallel, but halts actions like rollbacks until an engineer provides approval. Infrastructure costs remain low because MicroVMs bill only for storage when paused, resuming per-second execution only during active tasks. Development was completed using the LangChain Deep Agents framework, though a fully managed service is available for hands-off deployment.

Key points

  • Idle MicroVMs pause execution to bill only for storage during model inference or while aw…: Idle MicroVMs pause execution to bill only for storage during model inference or while awaiting operator decisions.
  • The platform uses Claude Sonnet 4.6 for commander duties and Claude Haiku 4.5 for the rem…: The platform uses Claude Sonnet 4.6 for commander duties and Claude Haiku 4.5 for the remaining support roles.
  • Organizations seeking to bypass manual configuration can deploy the automated architectur…: Organizations seeking to bypass manual configuration can deploy the automated architecture through DigitalOcean's Managed Agents service.

Sources checked by our analysis pipeline: 1

By Kushal Walia and Calin Lupa

A deploy goes out, and checkout starts failing. Within minutes, three agents are investigating in parallel: one reading metrics, one reading commits, and one checking the Droplet’s snapshots. Each works in its own MicroVM, none holding a credential for anything outside.

Then the team does something most agent demos skip: it stops and asks a person to approve the rollback. While that person decides, every MicroVM the investigation used pauses and bills only for storage.

We built this on-call team of AI agents on DigitalOcean MicroVMs. But the agents are only half of the story. What made the team safe to run, and cheap to leave waiting, was the infrastructure beneath it. This guide covers that first, then follows the incident from alert to verified fix. The code to run it yourself is at the end.

Once an agent starts doing production work, it requires real infrastructure:

When the work splits across subagents, each one needs all of these components, and fast. And because agents spend much of their time waiting, on a model’s…

Artificial Intelligence